From e2b67a868b7191237374226218756c1a62fabd4e Mon Sep 17 00:00:00 2001
From: ReinUsesLisp <reinuseslisp@airmail.cc>
Date: Fri, 15 May 2020 01:43:44 -0300
Subject: shader/other: Implement thread comparisons (NV_shader_thread_group)

Hardware S2R special registers match gl_Thread*MaskNV. We can trivially
implement these using Nvidia's extension on OpenGL or naively stubbing
them with the ARB instructions to match. This might cause issues if the
host device warp size doesn't match Nvidia's. That said, this is
unlikely on proper shaders.

Refer to the attached url for more documentation about these flags.
https://www.khronos.org/registry/OpenGL/extensions/NV/NV_shader_thread_group.txt
---
 .../renderer_opengl/gl_shader_decompiler.cpp       | 23 ++++++++++++++++++++++
 1 file changed, 23 insertions(+)

(limited to 'src/video_core/renderer_opengl')

diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index 960ebf1a1..c83a08d42 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -2309,6 +2309,18 @@ private:
         return {"gl_SubGroupInvocationARB", Type::Uint};
     }
 
+    template <const std::string_view& comparison>
+    Expression ThreadMask(Operation) {
+        if (device.HasWarpIntrinsics()) {
+            return {fmt::format("gl_Thread{}MaskNV", comparison), Type::Uint};
+        }
+        if (device.HasShaderBallot()) {
+            return {fmt::format("uint(gl_SubGroup{}MaskARB)", comparison), Type::Uint};
+        }
+        LOG_ERROR(Render_OpenGL, "Thread mask intrinsics are required by the shader");
+        return {"0U", Type::Uint};
+    }
+
     Expression ShuffleIndexed(Operation operation) {
         std::string value = VisitOperand(operation, 0).AsFloat();
 
@@ -2337,6 +2349,12 @@ private:
         static constexpr std::string_view NotEqual = "!=";
         static constexpr std::string_view GreaterEqual = ">=";
 
+        static constexpr std::string_view Eq = "Eq";
+        static constexpr std::string_view Ge = "Ge";
+        static constexpr std::string_view Gt = "Gt";
+        static constexpr std::string_view Le = "Le";
+        static constexpr std::string_view Lt = "Lt";
+
         static constexpr std::string_view Add = "Add";
         static constexpr std::string_view Min = "Min";
         static constexpr std::string_view Max = "Max";
@@ -2554,6 +2572,11 @@ private:
         &GLSLDecompiler::VoteEqual,
 
         &GLSLDecompiler::ThreadId,
+        &GLSLDecompiler::ThreadMask<Func::Eq>,
+        &GLSLDecompiler::ThreadMask<Func::Ge>,
+        &GLSLDecompiler::ThreadMask<Func::Gt>,
+        &GLSLDecompiler::ThreadMask<Func::Le>,
+        &GLSLDecompiler::ThreadMask<Func::Lt>,
         &GLSLDecompiler::ShuffleIndexed,
 
         &GLSLDecompiler::MemoryBarrierGL,
-- 
cgit v1.2.3